integrating reserve variabilty and erm · – model assumptions require validation and should...

Integrating Reserve Variability and ERM:A Case Study

1

www.actuarialsoftware.com

Mark R. Shapland, FCAS, FSA, MAAAJeffrey A. Courchene, FCAS, MAAA

Integrating Reserve Variability and ERM:

A Case Study

International Congress of Actuaries30 March – 4 April 2014

Washington, DC


What are the Issues?

How good are your estimates (mean, std. dev., etc.)? When will you know if your estimate is good? How do you compare actual outcomes to your estimate?

– How far apart and still reasonable?

Can you manage reserve risk:– Without measuring it first?– If the assumptions are not consistent over time?

Will retrospective testing improve your processes?– Are the inevitable deviations from the expectations understood?– Is there a difference between predicting & explaining?

What metrics are useful for management? Should we integrate reserving into ERM?

– Analysis of change, risk capital, earnings, etc.


Drivers of Change

International Accounting Standards (IFRS)

– Building Block, Risk Adjustment, Disclosure

Solvency II

– Quantification, Validation, Governance

NAIC Model Audit Rule

– Internal Data, Process, Reporting Validation

Own Risk Solvency Assessment (ORSA)

– Model Act Fall, 2012 Effective 1/1/15


2


Integrated ERM Framework

Conduct deterministic analysis to get a best estimate (BE) or central estimate

Conduct stochastic modeling of unpaid claim liabilities– Multiple models weighted to address model risk

Set threshold for action based on deviation from expected– Strategic allocation of actuarial talent during high pressure season

Automatically notify key personnel of unusual values at an early stage of the reserving process– Facilitate prompt investigation of potential data inaccuracies

– Make changes to the assumption set as needed, maintaining consistency of approach


Back Testing

Goal: Compare actual (A) to expected (E)

Deriving E requires assumption consistency

Assess materiality of difference (A - E)

– Expected (distributional) vs. Actual (one observation)

Caveats:

– Model assumptions require validation and should address model risk

– Does not address AY=CY. New exposures have been earned!

– Works well for gross but net (or R/I recoveries) requires more effort

– May need to “shift” mean of resulting distribution to replicate BE

0% 5% 25% 75% 95% 100%


Actual Expected Modeled Actual Expected Modeled

AY Age Paid Paid Percentile Incurred Incurred Percentile

2004 120 543 577 57.5% (47) 152 0.2%

2005 108 2,387 1,043 91.8% 1,040 503 81.9%

2006 96 1,177 1,636 35.6% 851 1,193 43.6%

2007 84 5,403 4,540 74.1% 2,954 2,064 79.5%

2008 72 14,120 10,630 93.5% 9,035 6,013 92.5%

2009 60 23,636 23,300 56.2% 16,524 11,898 95.0%

2010 48 51,020 44,746 88.8% 36,454 29,808 91.6%

2011 36 75,813 62,082 96.9% 61,541 44,977 99.0%

2012 24 88,832 79,335 87.0% 83,154 67,322 95.9%

2013 12 99,123 - 178,539 -

CY 2013 362,054 390,045

AY<CY 262,931 227,890 99.6% 211,506 163,930 99.9%

What can be measured without an uncertainty analysis?


3


Imagine the following…

The date is 2 January 2014

Complete loss data is available as of 31 December 2013

Company A writes 3 homogenous lines of business (CA, PPA, and HO), with triangular data going back to Accident Year 2004 (source: SNL Financial)

Company A performs a full review of unpaid claim liabilities annually, including an uncertainty analysis using multiple models to address model risk


Imagine the following…

Company A has an integrated risk management framework, including reserving risk Key Performance Indicators (KPIs), based on the realization of paid (and incurred) loss relative to outcomes of their models and pre-defined thresholds

Management would like to receive the actuary’s best estimate as of 31 December 2013 by 23 January 2014 (3 weeks)

0% 5% 25% 75% 95% 100%


Monitor/Control Reserving RiskCompare actual to expected (AY<CY)

450,000 500,000 550,000 600,000 650,000 700,000 750,000

180,000 200,000 220,000 240,000 260,000 280,000 300,000 100,000 120,000 140,000 160,000 180,000 200,000 220,000 240,000

0 50,000 100,000 150,000 200,000150,000 200,000 250,000 300,000 350,000 400,000

800,000 900,000 1,000,000 1,100,000 1,200,000

1,400,000 1,500,000 1,600,000 1,700,000 1,800,000 600,000 700,000 800,000 900,000 1,000,000

Aggregate Incurred Loss

PPA Incurred

CA Incurred

HO Incurred

Aggregate Paid Loss

PPA Paid

CA Paid

HO Paid


4




2004 120 3,069 3,672 35.4% 1,863 2,130 48.0%

2005 108 5,905 4,268 81.3% 3,145 1,751 81.6%

2006 96 8,986 10,276 32.3% 3,553 6,028 21.1%

2007 84 18,992 20,311 35.5% 9,872 9,977 52.2%

2008 72 51,003 49,291 64.6% 25,942 24,623 62.2%

2009 60 105,067 105,616 47.8% 52,012 51,904 52.8%

2010 48 202,932 197,620 69.1% 106,624 102,833 66.4%

2011 36 334,434 336,607 45.4% 189,908 179,363 76.8%

2012 24 841,484 845,014 47.7% 454,217 460,518 42.3%

2013 12 1,798,138 - 2,528,235 -

CY 2013 3,370,010 3,375,371

AY<CY 1,571,872 1,572,674 50.0% 847,136 839,128 59.1%

Monitor/Control Reserving RiskCompare actual to expected (AY<CY) Aggregate

Several of the 20 observable outcomes are near the thresholds

– 20 observable outcomes = (9 AYs + 1 AY<CY) for paid and incurred

AY 2013 could be addressed if pricing risk was included in analysis

NOTE:Comparison of aggregate accruals requires correlation assumptions


Integrated ERM FrameworkNon-Life Reserve Risk KPI: Observation (Aggregate)

No thresholds breached

Are we overestimating uncertainty?

Is the 80th percentile value surprising, given that we have 9 AY observations?


Integrated ERM FrameworkNon-Life Reserve Risk KPI: Aggregate Paid

Risk Owner

Risk Reviewer

Thresholds

Realized Values

AY / UY Details


5


Integrated ERM FrameworkAutomated E-Mail to the CFO


Monitor/Control Reserving RiskDo outcomes tell us something? (AY<CY)

Overall actual results are consistent with expectations– Includes both AY and Total (AY<CY) outcomes (20 outcomes each)

• Comparison of aggregate accruals requires correlation assumptions

– Includes both LoB and Aggregate outcomes (80 outcomes total)– CA could be problematic

• Internal process (data quality / claims adjusting / reinsurance)• Width of distribution or some other modeling assumption• Random occurrence

Number Percentage25<X<75 5<X<95 <5 or >95 25<X<75 5<X<95 <5 or >95

HO 10 13 18 20 2 0 50.0% 65.0% 90.0% 100.0% 10.0% 0.0%PPA 10 14 18 20 2 0 50.0% 70.0% 90.0% 100.0% 10.0% 0.0%CA 10 5 18 14 2 6 50.0% 25.0% 90.0% 70.0% 10.0% 30.0%Agg 10 16 18 20 2 0 50.0% 80.0% 90.0% 100.0% 10.0% 0.0%Total 40 48 72 74 8 6 50.0% 60.0% 90.0% 92.5% 10.0% 7.5%


Monitor/Control Reserving RiskOne-year time horizon reserve changes (AY<CY)

Given the actual losses paid in CY 2013, we can obtain a preliminary estimate of the amount by which reserves for AY 2012 and prior (or AY<CY) will change

– All the necessary information is contained within the prior deterministic analysis and uncertainty analysis (does not require an update with new data)

– Provides an early warning of impact on financial results

– Provides a measure of the performance of the actuarial function


6



Calculate, separately for each LOB:– “Conditional Reserve @ 31 December 2013” = Nth Percentile

• Example: If CY Paid fell into the 15th percentile of the distribution of expected CY Paid, the Conditional Reserve would be the 15th percentile of the distribution of reserves @ 31 December 2013

– “Expected Reserve @ 31 December 2013” = Expected Reserve @ 31 December 2012 less CY 2013 Paid

• This is the reserve @ 31 December 2013 if we did not change Ultimates at all

– Difference between Conditional Reserve and Expected Reserve represents the estimated reserve change

N

PointEstimates

(BE12+)

PossibleOutcomes(Sample Triangles)

N

Re-ParameterizeModel

(Sample Trapezoids)

N

Parameter/Process

Risk

PossibleOutcomes

(Future Outcomes – p1)


CA PPA HOExpected Conditional Expected Conditional Expected Conditional Total

AY Reserve Reserve Change Reserve Reserve Change Reserve Reserve Change Change2004 613 547 (67) 2,737 2,493 (245) 392 25 (367) (678) 2005 (146) 2,194 2,340 6,210 6,874 664 979 744 (235) 2,769 2006 2,500 1,533 (967) 9,566 8,940 (626) 1,559 1,511 (49) (1,642) 2007 3,205 4,927 1,722 19,331 17,337 (1,994) 2,013 114 (1,899) (2,171) 2008 5,828 12,825 6,997 36,672 33,136 (3,535) 2,897 4,499 1,602 5,064 2009 19,494 20,176 682 73,732 74,597 865 6,005 4,315 (1,690) (143) 2010 44,250 57,573 13,323 156,541 153,517 (3,024) 12,219 14,416 2,197 12,496 2011 80,777 113,108 32,331 319,636 303,909 (15,727) 25,577 22,449 (3,129) 13,475 2012 146,195 171,586 25,391 587,371 588,683 1,313 65,979 59,340 (6,639) 20,065 2013

AY<CY 302,716 384,469 81,754 1,211,797 1,189,486 (22,310) 117,621 107,412 (10,209) 49,234


AYs 2010-12 should also drive reserves up

– Most of this increase is driven by CA


Integrated ERM FrameworkAutomated E-Mail to the CEO/CFO


7


Monitor/Control Reserving Risk

Focus on Commercial Auto (CA)




2004 120 543 577 57.5% (47) 152 0.2%

2005 108 2,387 1,043 91.8% 1,040 503 81.9%

2006 96 1,177 1,636 35.6% 851 1,193 43.6%

2007 84 5,403 4,540 74.1% 2,954 2,064 79.5%

2008 72 14,120 10,630 93.5% 9,035 6,013 92.5%

2009 60 23,636 23,300 56.2% 16,524 11,898 95.0%

2010 48 51,020 44,746 88.8% 36,454 29,808 91.6%

2011 36 75,813 62,082 96.9% 61,541 44,977 99.0%

2012 24 88,832 79,335 87.0% 83,154 67,322 95.9%

2013 12 99,123 - 178,539 -

CY 2013 362,054 390,045

AY<CY 262,931 227,890 99.6% 211,506 163,930 99.9%

Monitor/Control Reserving RiskCompare CA actual to expected (AY<CY)

CA

AYs 2007-12 are driving high #s

– Need to check assumptions (i.e., IELRs, LDFs, weights, etc.)


Monitor/Control Reserving RiskCompare CA actual to expected (AY<CY)

CA Paid

AYs 2007-12 are driving high #s

– Need to check all assumptions

0 20,000 40,000 60,000 80,000 100,000 120,000 0 20,000 40,000 60,000 80,000 100,000 120,000

CA Incurred


8


Integrated ERM FrameworkNon-Life Reserve Risk KPI: Observation (LOB: CA)

Threshold breached Are expectations

from the 2012 model biased low?Check 2011

Are we aware of all internal process changes?

Are we underestimating uncertainty?


Integrated ERM FrameworkAutomated E-Mail to the Chief Actuary


Integrated ERM FrameworkNon-Life Reserve Risk KPI: CA Paid (AY<CY) Output

Risk Owner

Risk Reviewer

Thresholds

Realized Values

AY / UY Details


9


Integrated ERM FrameworkAutomated E-Mail to Data Quality Department


Integrated ERM FrameworkAutomated E-Mail to Claims Department


Integrated ERM FrameworkAutomated E-Mail to the Reinsurance Department


10


Assumption ConsistencyWe validated last year. Why so far off the mark?

Choice of 2012 IELR?

– Management: 52.9%

– Incurred CL: 57.7%

– Paid CL: 57.3%

Heteroscedasticity?

Shifting mean of distribution?

Missed CY trend?

Actual Expected Model

AY Age Paid Paid Percentile

2004 120 543 577 57.5%

2005 108 2,387 1,043 91.8%

2006 96 1,177 1,636 35.6%

2007 84 5,403 4,540 74.1%

2008 72 14,120 10,630 93.5%

2009 60 23,636 23,300 56.2%

2010 48 51,020 44,746 88.8%

2011 36 75,813 62,082 96.9%

2012 24 88,832 79,335 87.0%

2013 12 99,123 -

CY 2013 362,054

AY<CY 262,931 227,890 99.6%


Validation as of 31 December 2012Assumptions: Each requiring validation Long term average LDFs

– No validated reason to use shorter term averages (e.g., WA of last 5)

– In this example, model is 100% consistent with calculation of BE

• If deterministic analysis uses a “picker approach” (to reflect observable trends), need to validate each “pick” and consider shifting output of stochastic uncertainty model.

Accident year independence

IELRs used in the BF Method

Heteroecthesious data (i.e., non-uniform exposures)– We use symmetrical triangles (e.g., AY x AY)

– Exposures are complete (not at interim valuation date) and have not significantly changed over time (e.g., no rapid growth)


Validation as of 31 December 2012Assumptions: Each requiring validation

Heteroscedasticity– Residuals assumed to be identically distributed with a mean of zero

– Residuals by development period more variable than others?

Gamma used for Process Variance

Coefficient of Variation of the IELRs used in BF Method

Weighting of methods


11


Validation as of 31 December 2012Assumptions: CA Paid Loss Triangle

Year 12 24 36 48 60 72 84 96 108 1202004 77,401 140,425 189,316 223,326 243,182 250,182 254,305 256,672 257,689 258,2322005 76,085 142,122 193,196 224,406 246,220 257,226 263,698 264,871 267,2582006 79,850 139,041 181,905 209,366 228,012 237,792 240,300 241,4772007 80,323 144,482 192,134 227,723 249,165 259,339 264,7422008 83,919 152,487 203,761 245,150 270,525 284,6452009 82,001 151,768 201,189 245,541 269,1772010 91,514 170,696 240,652 291,6722011 103,957 177,709 253,5222012 105,547 194,3792013 99,123

LDF 1.810 1.359 1.189 1.095 1.042 1.018 1.006 1.007 1.002 1.002CDF 3.456 1.909 1.405 1.182 1.079 1.036 1.017 1.011 1.004 1.002

Assumption: E[c(w,d+1)|c(w,1),…,c(w,d)] = c(w,d) x F(d)

Corr. = 0.967 P-Value = 0.000 Int. P-Value = 0.004

0.0

50.0K

100.0K

150.0K

200.0K

250.0K

0.0 20.0K 40.0K 60.0K 80.0K 100.0K 120.0K

Cum. (24) vs. Cum. (12)


0.0

50.0K

100.0K

150.0K

200.0K

250.0K

300.0K

0.0 50.0K 100.0K 150.0K 200.0K

Cum. (36) vs. Cum. (24)



0.0

50.0K

100.0K

150.0K

200.0K

250.0K

300.0K

350.0K

0.0 50.0K 100.0K 150.0K 200.0K 250.0K 300.0K

Cum. (36) vs. Cum. (24)


0.0

50.0K

100.0K

150.0K

200.0K

250.0K

300.0K

0.0 50.0K 100.0K 150.0K 200.0K

Cum. (24) vs. Cum. (12)

Year 12 24 36 48 60 72 84 96 108 1202004 133,521 185,161 221,635 241,420 251,646 255,508 256,596 258,041 258,524 258,4772005 128,727 187,403 222,093 247,345 258,712 265,636 269,558 270,758 271,7982006 132,567 181,263 209,262 226,237 236,863 241,107 242,171 243,0222007 137,295 188,962 222,624 247,335 258,856 265,496 268,4502008 142,862 202,363 239,239 269,940 281,376 290,4112009 138,650 199,791 239,719 266,101 282,6252010 151,778 227,353 282,394 318,8482011 169,171 235,983 297,5242012 177,611 260,7652013 178,539

LDF 1.424 1.203 1.110 1.048 1.024 1.009 1.005 1.003 1.000 1.001CDF 2.075 1.457 1.211 1.091 1.041 1.017 1.008 1.004 1.001 1.001

Validation as of 31 December 2012Assumptions: CA Incurred Loss Triangle

Assumption: E[c(w,d+1)|c(w,1),…,c(w,d)] = c(w,d) x F(d)


Validation as of 31 December 2012Implied Expectations: Use of Paid and Incurred

Each method produces a different expectation of paid (incurred) loss.

The mean of the distribution used in the back test of paid (incurred) loss should be consistent with the paid (incurred) loss inherent in the selected ultimate.

Expected Paid Losses during CY 2013AY PCL ICL PBF IBF Weighted

2004 572 572 573 573 5722005 1,049 1,067 1,068 1,086 1,0582006 1,642 1,643 1,647 1,648 1,6432007 4,560 4,591 4,590 4,621 4,5762008 10,624 10,683 10,695 10,750 10,6542009 23,280 23,275 23,355 23,346 23,2782010 44,341 44,838 44,779 45,145 44,7762011 61,648 62,476 61,823 62,374 62,0982012 85,007 85,716 78,521 80,114 79,317

AY<CY 232,723 234,862 227,052 229,656 227,972

Expected Incurred Losses during CY 2013AY PCL ICL PBF IBF Weighted

2004 155 155 156 156 1552005 498 507 499 507 5032006 1,217 1,217 1,219 1,220 1,2172007 2,101 2,116 2,101 2,115 2,1082008 6,027 6,061 6,037 6,067 6,0442009 11,917 11,915 11,960 11,956 11,9162010 29,648 29,980 29,698 29,941 29,8172011 44,910 45,513 44,640 45,037 44,8392012 73,543 74,156 66,582 67,932 67,257

AY<CY 170,016 171,620 162,892 164,931 163,856

Note: The difference between the expectation from various models can be material for young AYs


12


Validation as of 31 December 2012Assumptions: AY Independence

Assumption: {c(i,1), …, c(i,n)} & {c(j,1), …, c(j,n)} are independent for i≠jCY LDFs Paid Loss

Small Large AY 24 / 12 36 / 24 48 / 36 60 / 48 72 / 60 84 / 72 96 / 84 108 / 961 0 2004 1.814 1.348 1.180 1.089 1.029 1.016 1.009 1.0040 2 2005 1.868 1.359 1.162 1.097 1.045 1.025 1.0042 1 2006 1.741 1.308 1.151 1.089 1.043 1.0114 0 2007 1.799 1.330 1.185 1.094 1.0413 2 2008 1.817 1.336 1.203 1.1041 3 2009 1.851 1.326 1.2201 5 2010 1.865 1.4104 3 2011 1.709

Median 1.816 1.336 1.182 1.094 1.042 1.016 1.007 1.004

CY LDFs Incurred LossSmall Large AY 24 / 12 36 / 24 48 / 36 60 / 48 72 / 60 84 / 72 96 / 84 108 / 96

1 0 2004 1.387 1.197 1.089 1.042 1.015 1.004 1.006 1.0020 2 2005 1.456 1.185 1.114 1.046 1.027 1.015 1.0042 0 2006 1.367 1.154 1.081 1.047 1.018 1.0043 1 2007 1.376 1.178 1.111 1.047 1.0263 1 2008 1.416 1.182 1.128 1.0422 4 2009 1.441 1.200 1.1101 6 2010 1.498 1.2424 2 2011 1.395

Median 1.406 1.185 1.111 1.046 1.022 1.004 1.005 1.002


Validation as of 31 December 2012Assumptions: Exposures

Exposures have grown slowly since 2006

Re-ran simulation with exposure-adjusted data; minimal impact

Actual Initial Initial Alternative Alternative

AY Age Paid Expected Percentile Expected Percentile

2004 120 543 577 57.5% 574 58.5%

2005 108 2,387 1,043 91.8% 1,060 91.5%

2006 96 1,177 1,636 35.6% 1,639 35.3%

2007 84 5,403 4,540 74.1% 4,554 73.2%

2008 72 14,120 10,630 93.5% 10,677 93.5%

2009 60 23,636 23,300 56.2% 23,272 56.0%

2010 48 51,020 44,746 88.8% 44,745 88.7%

2011 36 75,813 62,082 96.9% 61,999 96.8%

2012 24 88,832 79,335 87.0% 79,473 86.4%

2013 12 99,123 -

CY 2013 362,054

AY<CY 262,931 227,890 99.6% 227,994 99.5%

AY Exposures

2004 48,886

2005 47,449

2006 45,935

2007 46,224

2008 47,132

2009 47,358

2010 48,855

2011 50,167

2012 51,644

Total 433,650


Validation as of 31 December 2012Assumptions: CA Paid Loss Diagnostics

Does themodel explain all the trends?

Do you have only random noise left?

Are the variances all the same?


13


Validation as of 31 December 2012Assumptions: CA Paid Loss Diagnostics

All positive outliers could indicate skewness

Normality still good though

We can still check heteroscedasticity


Validation as of 31 December 2012Assumptions: Process Variance

Assumed a Gamma distribution

Switching to Normal distribution had minimal impact



2004 120 543 577 57.5% 577 47.0%

2005 108 2,387 1,043 91.8% 1,048 92.2%

2006 96 1,177 1,636 35.6% 1,652 32.1%

2007 84 5,403 4,540 74.1% 4,550 72.4%

2008 72 14,120 10,630 93.5% 10,622 93.6%

2009 60 23,636 23,300 56.2% 23,260 55.4%

2010 48 51,020 44,746 88.8% 44,694 89.1%

2011 36 75,813 62,082 96.9% 62,102 97.2%

2012 24 88,832 79,335 87.0% 79,251 87.3%

2013 12 99,123 -

CY 2013 362,054

AY<CY 262,931 227,890 99.6% 227,754 99.6%


Validation as of 31 December 2012Assumptions: CA BF and Weighting

BF models

– IELR consistent with BE

– CoV (IELR) = 8%

Weights identical to BE

Coefficient of VariationChain Ladder (Unshif ted) IELR BF (Unshif ted)

AY Paid Incurred CoV Paid Incurred

2004 55.9% 56.5% 8.0% 79.8% 78.6%2005 49.4% 48.9% 8.0% 57.0% 56.5%2006 38.0% 37.3% 8.0% 41.9% 42.1%2007 24.4% 24.3% 8.0% 26.9% 26.8%2008 16.1% 15.3% 8.0% 17.9% 17.6%2009 11.3% 10.1% 8.0% 13.2% 12.9%2010 8.1% 6.9% 8.0% 10.6% 10.0%2011 7.2% 6.2% 8.0% 9.6% 8.5%2012 7.6% 6.6% 8.0% 9.1% 7.9%

Total 4.9% 4.0% 5.3% 4.8%

In this case, the use of the BF

adds variability to the resulting

distribution


14


Validation as of 31 December 2012Assumptions: CA BF and Weighting (Alternative)

BF models

– IELR consistent with BE

– CoV (IELR) = 0%

Weights identical to BE

Coefficient of VariationChain Ladder (Unshif ted) IELR BF (Unshif ted)

AY Paid Incurred CoV Paid Incurred

2004 55.9% 56.5% 0.0% 78.1% 78.5%2005 49.4% 48.9% 0.0% 56.0% 56.5%2006 38.0% 37.3% 0.0% 40.5% 40.9%2007 24.4% 24.3% 0.0% 25.7% 25.0%2008 16.1% 15.3% 0.0% 16.1% 15.9%2009 11.3% 10.1% 0.0% 10.4% 10.4%2010 8.1% 6.9% 0.0% 6.9% 7.0%2011 7.2% 6.2% 0.0% 5.1% 5.5%2012 7.6% 6.6% 0.0% 4.0% 4.7%

Total 4.9% 4.0% 3.1% 3.2%

In this case, the use of the BF

reduces variability of the

resulting distribution


Validation as of 31 December 2012Assumptions: CA IELR (for BF) and Weights

Optimism Regarding AY 2012 ULR

– In this example, IELR based on published figures (selected ultimate)

– IELR is an important assumption which requires additional validation

• Consider renewal study performed by Underwriting

• Consider actuarial analysis of average rate achieved

– Sensitivity tests confirm that this assumption is only a partial explanation

Paid CL Incurred CL Management SelectedAY ULR ULR IELR ULR

2004 73.2% 73.2% 73.3% 73.2%2005 76.0% 77.3% 77.4% 76.7%2006 64.5% 64.5% 64.6% 64.5%2007 62.8% 63.2% 63.2% 63.0%2008 60.4% 60.7% 60.8% 60.6%2009 53.2% 53.2% 53.4% 53.2%2010 57.9% 58.5% 58.5% 58.2%2011 54.5% 55.3% 54.7% 54.9%2012 57.3% 57.7% 52.9% 54.7%

Paid Incurred Incurred IncurredAY CL CL BF BF

2004 50.0% 50.0%2005 50.0% 50.0%2006 50.0% 50.0%2007 50.0% 50.0%2008 50.0% 50.0%2009 50.0% 50.0%2010 25.0% 25.0% 25.0% 25.0%2011 50.0% 50.0%2012 50.0% 50.0%


Assumption ConsistencyWe validated last year. Why so far off?

2012 IELR

– No longer 52.9%

– Used 57.5%

Explains AY 2012 deviation only.

Still breach LoBthreshold



2004 120 543 577 57.5% 566 57.8%

2005 108 2,387 1,043 91.8% 1,064 91.4%

2006 96 1,177 1,636 35.6% 1,639 35.2%

2007 84 5,403 4,540 74.1% 4,569 73.3%

2008 72 14,120 10,630 93.5% 10,650 93.1%

2009 60 23,636 23,300 56.2% 23,359 54.8%

2010 48 51,020 44,746 88.8% 44,662 89.3%

2011 36 75,813 62,082 96.9% 62,032 97.1%

2012 24 88,832 79,335 87.0% 85,452 66.2%

2013 12 99,123 -

CY 2013 362,054

AY<CY 262,931 227,890 99.6% 233,994 98.5%

IELR


15




2004 120 543 577 57.5% 574 61.4%

2005 108 2,387 1,043 91.8% 1,051 88.3%

2006 96 1,177 1,636 35.6% 1,646 37.4%

2007 84 5,403 4,540 74.1% 4,544 73.0%

2008 72 14,120 10,630 93.5% 10,664 91.1%

2009 60 23,636 23,300 56.2% 23,228 56.7%

2010 48 51,020 44,746 88.8% 44,751 82.9%

2011 36 75,813 62,082 96.9% 62,034 96.5%

2012 24 88,832 79,335 87.0% 79,373 89.1%

2013 12 99,123 -

CY 2013 362,054

AY<CY 262,931 227,890 99.6% 227,864 99.3%


Minimal impact

Still breach LoBthresholds

Heteroscedasticity


Assumption ConsistencyWe validated last year. Why so far off? CY Trend


New GLM model with CY Trend:1.9% Trend for 2004-2009 and 3.6% for 2009-2012+


16


Monitor/Control Reserving RiskImpact of change in prior assumption (AY<CY)

Adding CY trend parameter to model improves fit & results?– GLM model also adjusted for exposures

– Statistics comparable, some better, some not as good

ODP Paid Model GLM Paid ModelActual Expected Bootstrap Expected Bootstrap

AY Age Paid Paid Percentile Paid Percentile2004 120 543 577 57.5% 62 96.1%2005 108 2,387 1,043 91.8% 2,021 65.2%2006 96 1,177 1,636 35.6% 2,868 12.6%2007 84 5,403 4,540 74.1% 6,989 25.3%2008 72 14,120 10,630 93.5% 14,810 43.8%2009 60 23,636 23,300 56.2% 26,680 23.4%2010 48 51,020 44,746 88.8% 49,173 63.1%2011 36 75,813 62,082 96.9% 64,678 94.5%2012 24 88,832 79,335 87.0% 87,876 55.5%2013 12 99,123

CY 2013 362,054 AY<CY 262,931 227,890 99.6% 255,155 68.5%


Integrated ERM FrameworkManual E-Mail to the Claims Officer


Mack Model CalculationsStandard Expected Std Dev Actual

AY Reserve Deviation CoV Paid CY 13 CY 2013 Paid Percentile

2004 1,146 188 16.4% 1,146 188 543 0.0%2005 2,232 644 28.9% 1,049 615 2,387 96.3%2006 3,681 1,207 32.8% 1,642 1,046 1,177 39.0%2007 8,603 2,548 29.6% 4,560 2,199 5,403 72.6%2008 19,950 3,441 17.2% 10,624 2,152 14,120 93.6%2009 43,104 3,838 8.9% 23,280 1,727 23,636 59.6%2010 94,371 8,325 8.8% 44,341 7,177 51,020 83.0%2011 155,511 11,761 7.6% 61,648 8,335 75,813 94.6%2012 251,758 16,702 6.6% 85,007 11,349 88,832 65.5%

Total 580,356 26,820 4.6% 233,297 19,185 262,931 93.3%


Similar to using a “Shifted” paid Chain Ladder

Often seen in industry, but under this scenario:– Management‘s low 2012 IELR may not get attention

– Recent CY trends may not get attention

Mack Model

Must decompose

Mack formula and make

distribution assumption

Variance assumptions disconnected

from BE assumptions


17


Validation as of 31 December 2012Assumptions: Correlation by Segment

Measurement:– Use of rank or pairwise correlation of

paid residuals

– Could have used incurred residuals

Evaluation:– P-value is the probability of obtaining

a test statistic at least as extreme as the one that was actually observed, assuming that the null hypothesis is true.

– Could have used incurred residuals

– Could have used residuals after heteroscedasticity adjustment

– Can validate by tracking over time

In this case, the calculated

correlation is not significantly

different from zero.


Any Final Questions?

Mark R. Shapland, FCAS, FSA, MAAA18119 Bent Ridge DriveWildwood, MO 63038 USATel: +1 636 273 6428Mobile: +1 636 346 [email protected]

Jeffrey A. Courchene, FCAS, MAAAAltheimer Eck 280331 Munich, GermanyTel: + 49 89 127 108 712Mobile: + 49 160 554 [email protected]

integrating reserve variabilty and erm · – model assumptions require validation and should...

Documents